Sandboxing
Sandboxing is the isolation of software execution in a restricted environment that limits access to resources and effects on surrounding systems.
AI Governance
Sandboxing is the execution of software in a constrained environment that limits its access to systems, data, networks, and privileges.
- How it is used
- Contains agent workloads to limit damage when attacks bypass other controls.
- How deep it goes
- Specifies network, filesystem, process, and credential boundaries, and discusses configuration trade-offs, escapes, and latency.
- Examples
- Uses a claims agent in an isolated container with restricted database access, a temporary workspace, and no process spawning.
- What you can do after
- Operate/configure agent sandboxes by choosing access boundaries and weighing legitimate workload needs against containment.
Vibe Engineering
The isolation of software execution within constrained resources and permissions to limit its impact on the surrounding system.
- How it is used
- Limits the blast radius of agent mistakes before work reaches systems that matter.
- How deep it goes
- Explains isolation options and human-approved export, while stopping short of a detailed execution-sandbox setup.
- Examples
- Uses a separate modernization repository and a Conda environment to illustrate distinct forms of isolation.
- What you can do after
- Operate/configure a Conda environment and prepare a separate repository for AI-assisted refactoring.
Claude at Work
A sandbox environment is an isolated execution environment that restricts code and tool access from the surrounding system.
- How it is used
- Enables AI tools to execute commands and work with files within cloud or permission-controlled local boundaries.
- How deep it goes
- Explains where execution happens and how folder permissions affect reads and writes, without detailing sandbox internals.
- Examples
- Uses a Downloads folder inventory and a ZIP-based CSV analysis that produces a local report.
- What you can do after
- Apply it via Claude Desktop Cowork to grant folder access, run commands, and create local files.
Refactoring to Rust
A sandboxed environment is an isolated execution environment that restricts a program's access to resources outside explicitly permitted interfaces.
- How it is used
- Uses sandboxing to run replaceable Wasm modules within a host while limiting their access to the surrounding system.
- How deep it goes
- Explains isolation and walks through allocation, writing, execution, and reading across the module boundary.
- Examples
- Uses Rust search modules and an allocation function to illustrate host-to-module memory exchange.
- What you can do after
- Evaluate/understand how Wasm isolation limits module access and why the VM remains responsible for handling exchanged data.
Build an AI Agent (From Scratch)
A sandbox environment is an isolated execution environment that restricts code or tool operations from affecting the surrounding system.
- How it is used
- Enables agents to execute generated code in isolation rather than relying only on predefined tool calls.
- How deep it goes
- Explains isolation risks and protections, then demonstrates sandbox lifecycle, configuration, state persistence, and execution errors.
- Examples
- Uses a persistent calculation, Wikipedia revision counts, and spreadsheet analysis to illustrate sandboxed execution.
- What you can do after
- Apply it via E2B to create a sandbox, execute Python, inspect results, and manage its lifetime.
Read how Build an AI Agent (From Scratch) explains sandboxing →
Context Engineering
Execution of generated code or commands within a restricted environment that limits their access and effects.
- How it is used
- Uses sandboxing to isolate agent execution while keeping live environment state outside the context window.
- How deep it goes
- Explains the command-and-feedback flow and its safety role without specifying a sandbox implementation.
- Examples
- Illustrates a persistent dataset whose schema and row counts, rather than its full contents, enter the prompt.
- What you can do after
- Evaluate sandboxing as both a context-management boundary and a safety control alongside other runtime safeguards.
Fundamentals of Cybersecurity
Sandboxing is the execution of potentially unsafe files or programs in a controlled environment isolated from operational systems.
- How it is used
- Uses isolated execution to detect and safely analyze suspicious files as part of endpoint protection.
- How deep it goes
- Explains isolation and contained behaviors through a scenario, without specifying sandbox setup or configuration.
- Examples
- Demonstrates a suspicious email attachment attempting malware installation, sensitive-file changes, and external connections inside a sandbox.
- What you can do after
- Evaluate suspicious files by observing their behavior in isolation before they affect devices or networks.
Read how Fundamentals of Cybersecurity explains sandboxing →
Server-Side WebAssembly
Sandboxing is a security technique that confines code to a restricted execution environment and limits its access to host resources.
- How it is used
- Uses restricted execution to isolate Wasm components, protect cloud applications, and safely run application plugins.
- How deep it goes
- Explains memory-safe isolation, capability-controlled OS access, and supply-chain benefits, while stopping short of implementation commands or runtime configuration.
- Examples
- Uses controlled standard output, isolated cloud components, application plugins, and comparisons with containers and virtual machines.
- What you can do after
- Evaluate Wasm as an isolation mechanism for plugins and cloud components while recognizing multithreading and ecosystem limitations.
AI Model Evaluation
An isolated execution environment that enables experimentation without affecting real users, systems, or production traffic.
- How it is used
- Provides a safe, controlled setting for evaluating model behavior before changes affect production traffic or user experience.
- How deep it goes
- Defines isolation and outlines historical-data testing, metric analysis, drift diagnosis, and bias investigation without implementation internals.
- Examples
- Uses historical interaction logs, labeled datasets, on-demand model invocations, model-version comparisons, and metric analysis.
- What you can do after
- Evaluate/understand model changes, investigate failures, and explore product optimizations without risking real users or systems.
Designing AI Agents
A sandbox is an isolated execution environment that restricts a system's access to resources and external effects.
- How it is used
- Constrains agent actions so routine work can proceed within known boundaries while approval policies handle actions requiring review.
- How deep it goes
- Compares isolation boundaries and implements policy checks, but delegates physical isolation and some resource enforcement to a runtime adapter.
- Examples
- Uses Claude Code, Codex, E2B, Manus, and a Python executor with path, rate, cost, and runtime limits as examples.
- What you can do after
- Configure a policy envelope and compose checks before invoking a runtime that enforces actual isolation.
Also covered in
- Grokking Simplicity
- AI Agents in Action, Second Edition
- Grokking Web Application Security
- Architecting Autonomous Agentic SystemsIn early access (MEAP)
- The Cloud at Your Service: The when, how, and why of enterprise cloud computing
- Introduction to Generative AI, Second Edition
- Think Like a CTO
- Vanilla WebIn early access (MEAP)
- AI Agents for Offensive SecurityIn early access (MEAP)
- Building Reliable AI Systems
- Effective Platform Engineering
What to know first
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- PythonPython is a high-level, general-purpose, dynamically typed programming language with readable syntax and support for multiple programming paradigms.
- containerA software container is an isolated execution environment that packages an application and its dependencies, typically sharing the host operating-system kernel.
- prompt injectionPrompt injection is an attack that introduces instructions through input or context to manipulate an AI system’s behavior or override its intended instructions.
Related concepts
- Claude Codecontrasts with · Claude Code is Anthropic’s command-line AI coding agent for inspecting and modifying codebases and carrying out software-development tasks through natural-language interaction and tools.
- execution environmentkind of · An execution environment is the computational context in which a program runs, including its supporting software, available resources, and relevant system state or configuration.
- browserimplemented by · A web browser is software that retrieves, interprets, and renders web resources, executes supported web content, and enables interaction with web applications.
- blast radiusaddresses · Blast radius is the scope of systems, components, data, users, or functionality that may be affected by a failure, change, or other disruptive event.
- code executionuses · Code execution is the process by which a computing system runs program instructions to produce results or effects.
Questions about sandboxing
What is sandboxing?
Sandboxing is the isolation of software execution in a restricted environment that limits access to resources and effects on surrounding systems.
Which Manning books explain sandboxing?
AI Governance, Vibe Engineering, Claude at Work, Refactoring to Rust, Build an AI Agent (From Scratch), Context Engineering, Fundamentals of Cybersecurity, Server-Side WebAssembly and 2 more.
What should I know before learning sandboxing?
Manning books that teach sandboxing usually introduce language model, Python, container and prompt injection first.
What is the difference between sandboxing and Claude Code?
Sandboxing is the isolation of software execution in a restricted environment that limits access to resources and effects on surrounding systems. Claude Code is Anthropic’s command-line AI coding agent for inspecting and modifying codebases and carrying out software-development tasks through natural-language interaction and tools.
What is sandboxing a kind of?
Sandboxing is a kind of execution environment. An execution environment is the computational context in which a program runs, including its supporting software, available resources, and relevant system state or configuration.
What is sandboxing also called?
Sandboxing is also called Sandbox or sandboxed execution.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.









